September 21, :43pm Holm Vol 2 WSPC/Book Trim Size for 9in by 6in

Transcription

1 1 GALILEO Contents 1.1 Principle of Galilean relativity Galilean transformations Admissible force laws for an N-particle system Subgroups of the Galilean transformations Matrix representation of SE(3) Lie group actions of SE(3) Lie group actions of G(3) Matrix representation of G(3) Lie algebra of SE(3) Lie algebra of G(3) 17

2 2 1 : GALILEO 1.1 Principle of Galilean relativity Galileo Galilei Principles of relativity address the problem of how events that occur in one place or state of motion are observed from another. And if events occurring in one place or state of motion look different from those in another, how should one determine the laws of motion? Galileo approached this problem via a thought experiment which imagined observations of motion made inside a ship by people who could not see outside. He showed that the people isolated inside a uniformly moving ship would be unable to determine by measurements made inside it whether they were moving!... have the ship proceed with any speed you like, so long as the motion is uniform and not fluctuating this way and that. You will discover not the least change in all the effects named, nor could you tell from any of them whether the ship was moving or standing still. Galileo, Dialogue Concerning the Two Chief World Systems [Ga1632] Galileo s thought experiment showed that a man who is below decks on a ship cannot tell whether the ship is docked or is moving uniformly through the water at constant velocity. He may observe water dripping from a bottle, fish swimming in a tank, butterflies flying, etc. Their behaviour will be just the same, whether the ship is moving or not. Definition (Galilean transformations) Transformations of reference location, time, orientation or state of uniform translation at constant velocity are called Galilean transformations.

3 1.2 GALILEAN TRANSFORMATIONS 3 Definition (Uniform rectilinear motion) Coordinate systems related by Galilean transformations are said to be in uniform rectilinear motion relative to each other. Galileo s thought experiment led him to the following principle. Definition (Principle of Galilean relativity) The laws of motion are independent of reference location, time, orientation or state of uniform translation at constant velocity. Hence, these laws are invariant (i.e., they do not change their forms) under Galilean transformations. Remark (Two tenets of Galilean relativity) Galilean relativity sets out two important tenets: It is impossible to determine who is actually at rest. Objects continue in uniform motion unless acted upon. The second tenet is known as Galileo s law of inertia. It is also the basis for Newton s first law of motion. 1.2 Galilean transformations Definition (Galilean transformations) Galilean transformations of a coordinate frame consist of space-time translations, rotations and reflections of spatial coordinates, as well as Galilean boosts into uniform rectilinear motion. In three dimensions, the Galilean transformations depend smoothly on ten real parameters, as follows: Space-time translations, g 1 (r, t) = (r + r 0, t + t 0 ). These possess four real parameters: (r 0, t 0 ) R 3 R, for the three dimensions of space, plus time.

4 2 NEWTON, LAGRANGE, HAMILTON AND THE RIGID BODY Contents 2.1 Newton Newtonian form of free rigid rotation Newtonian form of rigid-body motion Lagrange The principle of stationary action Noether s theorem Lie symmetries and conservation laws Infinitesimal transformations of a Lie group Lagrangian form of rigid-body motion Hamilton Pontryagin constrained variations Manakov s formulation of the SO(n) rigid body Matrix Euler Poincaré equations An isospectral eigenvalue problem for the SO(n) rigid body 56

5 20 2 : NEWTON, LAGRANGE, HAMILTON Manakov s integration of the SO(n) rigid body Hamilton Hamiltonian form of rigid-body motion Lie Poisson Hamiltonian rigid-body dynamics Lie Poisson bracket Nambu s R 3 Poisson bracket Clebsch variational principle for the rigid body Rotating motion with potential energy 72

6 2.1 NEWTON Newton Newtonian form of free rigid rotation Definition In free rigid rotation a body rotates about its centre of mass and the pairwise distances between all points in the body remain fixed. Definition A system of coordinates fixed in a body undergoing free rigid rotation is stationary in the rotating orthonormal basis called the body frame, introduced by Euler [Eu1758]. The orientation of the orthonormal frame (E 1, E 2, E 3 ) fixed in the rotating body relative to a basis (e 1, e 2, e 3 ) Isaac Newton fixed in space depends smoothly on time t R. In the fixed spatial coordinate system, the body frame is seen as the moving frame (O(t)E 1, O(t)E 2, O(t)E 3 ), where O(t) SO(3) defines the attitude of the body relative to its reference configuration according to the following matrix multiplication on its three unit vectors: e a (t) = O(t)E a, a = 1, 2, 3. (2.1.1) Here the unit vectors e a (0) = E a with a = 1, 2, 3 comprise at initial time t = 0 an orthonormal basis of coordinates and O(t) is a special (det O(t) = 1) orthogonal (O T (t)o(t) = Id) 3 3 matrix. That is, O(t) is a continuous function defined along a curve parameterised by time t in the special orthogonal matrix group SO(3). At the initial time t = 0, we may take O(0) = Id, without any loss.

7 2.1 NEWTON 29 Proof. The time derivative of relation (2.1.18) gives dj body dt = d ( ) O 1 (t)j space (t) dt ( ) = O 1 Ȯ O 1 (t)j space (t) + O 1 dj space }{{ dt } vanishes = ω body J body = ω body J body, (2.1.20) with ω body := O 1 Ȯ = ω body and ω body = I 1 J body, which defines the body angular velocity. This is the usual heuristic derivation of the dynamics for body angular momentum. Remark (Darwin, Coriolis and centrifugal forces) Many elementary mechanics texts make the following points about the various noninertial forces that arise in a rotating frame. For any vector r(t) = r a (t)e a (t) the body and space time derivatives satisfy the first time-derivative relation, as in (2.1.17), ṙ(t) = ṙ a e a (t) + r a ė a (t) = ṙ a e a (t) + ω r a e a (t) = ( ṙ a + ɛ a bc ωb r c) e a (t) =: (ṙa + (ω r) a) e a (t). (2.1.21) Taking a second time derivative in this notation yields r(t) = ( r a + ( ω r) a + (ω ṙ) a) e a (t) + ( ṙ a + (ω r) a) ė a (t) = ( r a + ( ω r) a + (ω ṙ) a) e a (t) + ( ω ( ṙ + ω r )) a ea (t) = ( r a + ( ω r) a + 2(ω ṙ) a + (ω ω r) ) a ea (t). Newton s second law for the evolution of the position vector r(t) of a particle of mass m in a frame rotating with time-dependent angular velocity ω(t) becomes F(r) = m ( r ) + ω r + 2(ω ṙ) + ω (ω r). (2.1.22) }{{}}{{}}{{} Darwin Coriolis centrifugal

8 30 2 : NEWTON, LAGRANGE, HAMILTON The Darwin force is usually small; so it is often neglected. Only the Coriolis force depends on the velocity in the moving frame. The Coriolis force is very important in largescale motions on Earth. For example, pressure balance with the Coriolis force dominates the (geostrophic) motion of weather systems that comprise the climate. The centrifugal force is important, for example, in obtaining orbital equilibria in gravitationally attracting systems. Remark The space and body angular velocities differ by ω body := O 1 Ȯ versus ω space := ȮO 1 = O ω body O 1. Namely, ω body is left-invariant under O RO and ω space is rightinvariant under O OR, for any choice of matrix R SO(3). This means that neither angular velocity depends on the initial orientation. Remark The angular velocities ω body = O 1 Ȯ and ω space = ȮO 1 are respectively the left and right translations to the identity of the tangent matrix Ȯ(t) at O(t). These are called the left and right tangent spaces of SO(3) at its identity. Remark Equations (2.1.19) for free rigid rotations of particle systems are prototypes of Euler s equations for the motion of a rigid body Newtonian form of rigid-body motion In describing rotations of a rigid body, for example a solid object occupying a spatial domain B R 3, one replaces the mass-weighted sums over points in space in the previous definitions of dynamical

9 2.3 NOETHER S THEOREM Noether s theorem Lie symmetries and conservation laws Recall from Definition that a Lie group depends smoothly on its parameters. (See Appendix B for more details.) Definition (Lie symmetry) A smooth transformation of variables {t, q} depending on a single parameter s defined by {t, q} { t(t, q, s), q(t, q, s)}, Emmy Noether that leaves the action S = L dt invariant is called a Lie symmetry of the action. Theorem (Noether s theorem) Each Lie symmetry of the action for a Lagrangian system defined on a manifold M with Lagrangian L corresponds to a constant of the motion [No1918]. Example Suppose the variation of the action in (2.2.3) vanishes (δs = 0) because of a Lie symmetry which does not preserve the endpoints. Then on solutions of the Euler Lagrange equations, the endpoint term must vanish for another reason. For example, if the Lie symmetry leaves time invariant, so that then the endpoint term must vanish, {t, q} { t, q(t, q, s)}, [ ] L t2 q δq = 0. t 1 Hence, the quantity A(q, q, δq) = L q a δqa

10 40 2 : NEWTON, LAGRANGE, HAMILTON is a constant of motion for solutions of the Euler Lagrange equations. In particular, if δq a = c a for constants c a, a = 1,..., n, that is, for spatial translations in n dimensions, then the quantities L/ q a (the corresponding momentum components) are constants of motion. Remark This result first appeared in Noether [No1918]. See, e.g., [Ol2000, SaCa1981] for good discussions of the history, framework and applications of Noether s theorem. We shall see in a moment that Lie symmetries that reparameterise time may also yield constants of motion Infinitesimal transformations of a Lie group Definition (Infinitesimal Lie transformations) Consider the Lie group of transformations {t, q} { t(t, q, s), q(t, q, s)}, Sophus Lie and suppose the identity transformation is arranged to occur for s = 0. The derivatives with respect to the group parameters s at the identity, τ(t, q) = d ds t(t, q, s), s=0 ξ a (t, q) = d ds q a (t, q, s), s=0 are called the infinitesimal transformations of the action of a Lie group on the time and space variables. Thus, at linear order in a Taylor expansion in the group parameter s one has t = t + sτ(t, q), q a = q a + sξ a (t, q), (2.3.1) where τ and ξ a are functions of coordinates and time, but do not depend on velocities. Then, to first order in s the velocities of the

11 2.4 LAGRANGIAN FORM OF RIGID-BODY MOTION 49 Thus, Euler s equations for the rigid body in T R 3, I Ω = IΩ Ω, (2.4.8) do follow from the variational principle (2.4.3) with variations of the form (2.4.5) derived from the definition of body angular velocity Ω. Remark The body angular velocity is expressed in terms of the spatial angular velocity by Ω(t) = O 1 (t)ω(t). Consequently, the kinetic energy Lagrangian in (2.4.4) transforms as l(ω) = 1 2 Ω IΩ = 1 2 ω I space(t)ω =: l space (ω), where as in (2.1.31). I space (t) = O(t)IO 1 (t), Exercise. Show that Hamilton s principle for the action S(ω) = b a l space (ω) dt yields conservation of spatial angular momentum π = I space (t)ω(t). Hint: First derive δi space = [ξ, I space ] with right-invariant ξ = δoo 1. Exercise. (Noether s theorem for the rigid body) What conservation law does Theorem (Noether s theorem) imply for the rigid-body Equations (2.4.2)?

12 50 2 : NEWTON, LAGRANGE, HAMILTON Hint: Transform the endpoint terms arising on integrating the variation δs by parts in the proof of Theorem into the spatial representation by setting Ξ = O 1 (t)γ and Ω = O 1 (t)ω. Remark (Reconstruction of O(t) SO(3)) The Euler solution is expressed in terms of the time-dependent angular velocity vector in the body, Ω. The body angular velocity vector Ω(t) yields the tangent vector Ȯ(t) T O(t)SO(3) along the integral curve in the rotation group O(t) SO(3) by the relation Ȯ(t) = O(t) Ω(t), (2.4.9) where the left-invariant skew-symmetric 3 3 matrix Ω is defined by the hat map (2.1.27) (O 1 Ȯ) jk = Ω jk = Ω i ɛ ijk. (2.4.10) Equation (2.4.9) is the reconstruction formula for O(t) SO(3). Once the time dependence of Ω(t) and hence Ω(t) is determined from the Euler equations, solving formula (2.4.9) as a linear differential equation with time-dependent coefficients yields the integral curve O(t) SO(3) for the orientation of the rigid body Hamilton Pontryagin constrained variations Formula (2.4.6) for the variation Ω of the skew-symmetric matrix Ω = O 1 Ȯ

13 56 2 : NEWTON, LAGRANGE, HAMILTON Proof. The proof here is the same as the proof of Manakov s commutator formula via Hamilton s principle, modulo replacing O 1 Ȯ so(n) with g 1 ġ g. Remark Poincaré s observation leading to the matrix Euler Poincaré Equation (2.4.27) was reported in two pages with no references [Po1901]. The proof above shows that the matrix Euler Poincaré equations possess a natural variational principle. Note that if Ω = g 1 ġ g, then M = δl/δω g, where the dual is defined in terms of the matrix trace pairing. Exercise. Retrace the proof of the variational principle for the Euler Poincaré equation, replacing the left-invariant quantity g 1 ġ with the right-invariant quantity ġg An isospectral eigenvalue problem for the SO(n) rigid body The solution of the SO(n) rigid-body dynamics dm dt = [ M, Ω ] with M = AΩ + ΩA, for the evolution of the n n skew-symmetric matrices M, Ω, with constant symmetric A, is given by a similarity transformation (later to be identified as coadjoint motion), M(t) = O(t) 1 M(0)O(t) =: Ad O(t) M(0), with O(t) SO(n) and Ω := O 1 Ȯ(t). Consequently, the evolution of M(t) is isospectral. This means that

14 2.4 LAGRANGIAN FORM OF RIGID-BODY MOTION 57 The initial eigenvalues of the matrix M(0) are preserved by the motion; that is, dλ/dt = 0 in M(t)ψ(t) = λψ(t), provided its eigenvectors ψ R n evolve according to ψ(t) = O(t) 1 ψ(0). The proof of this statement follows from the corresponding property of similarity transformations. Its matrix invariants are preserved: d dt tr(m λid)k = 0, for every non-negative integer power K. This is clear because the invariants of the matrix M may be expressed in terms of its eigenvalues; but these are invariant under a similarity transformation. Proposition Isospectrality allows the quadratic rigid-body dynamics (2.4.26) on SO(n) to be rephrased as a system of two coupled linear equations: the eigenvalue problem for M and an evolution equation for its eigenvectors ψ, as follows: Mψ = λψ and ψ = Ωψ, with Ω = O 1 Ȯ(t). Proof. Applying isospectrality in the time derivative of the first equation yields ( Ṁ + [ Ω, M ] )ψ + (M λid)( ψ + Ωψ) = 0. Now substitute the second equation to recover (2.4.26).